sonrai-en.html

<h2>Developing patterns for Irish</h2>
<p>
There is a brief description of the TeX
hyphenation algorithm on the
<a target="_blank" href="https://en.wikipedia.org/wiki/TeX#Hyphenation_and_justification">Wikipedia page</a> for TeX, 
or for more details you can check out
<a target="_blank" href="https://tug.org/docs/liang/liang-thesis.pdf">Frank Liang's PhD thesis</a> where the algorithm was first described.
</p>

<p>
The Irish patterns consist of rules like the following:
<pre>
al3i
a6ll
al2ann
geal5a
</pre>
Roughly speaking, even numbers prevent a word from being broken
at the given point, and odd numbers permit a break at the given point.
Larger numbers carry stronger weight than lower numbers when two rules
apply. The first rule <code>al3i</code> permits a hyphen after the “l”
in words like <i>béaliata</i> or <i>galinneall</i>.
The second rule strongly prevents hyphenation before
the “ll” in words like <i>timpeallacht</i> or <i>fealltóir</i>.
The third rule weakly prevents hyphenation after the “l” in 
words like <i>bialann</i> or <i>dialann</i>.  Note that it also applies
to the verb <i>gealann</i>, theoretically preventing a desirable
hyphenation, but is overridden by the fourth rule
which permits a hyphen at this same spot (since 5 is greater than 2).
</p>

<p>
One basic heuristic involves lenition;
a word should never be broken between a lenitable
consonant and the “h” indicating the lenition orthographically.
Thus you will find patterns like <code>c2h</code> and <code>d2h</code>
in the rule set.  Conversely, if an “h” appears after a vowel
or non-lenitable consonant, it is usually a good candidate for a hyphen point, 
as in <i lang="ga">Bói-héam-ach</i> or <i lang="ga">Faran-haít</i>.
This results in patterns
of the form <code>i1h</code> and <code>n5h6a</code>.
</p>

<p>
Another basic heuristic is, for syncopated words, to include
a hyphen at the point of syncopation; e.g. 
<i lang="ga">ciog-al</i> and <i lang="ga">ciog-lach</i>.
</p>

<hr>
<h2>Results</h2>
<p>
The resulting hyphenation patterns are very much morphological
vs. phonological (my personal preference).
As a consequence, they do not always
agree with hyphenations I've found in actual printed texts.
For instance:
</p>

<table lang="ga" class="fleiscin">
<tr><th>These patterns</th>  <th>Corpus</th></tr>
<tr><td>Ceilt-each</td>         <td>Ceil-teach</td></tr>
<tr><td>siosc-adh</td>          <td>sios-cadh</td></tr>
<tr><td>craic-eann</td>         <td>crai-ceann</td></tr>
<tr><td>ceann-aithe</td>        <td>cean-naithe</td></tr>
<tr><td>tuairt-eáil</td>        <td>tuair-teáil</td></tr>
<tr><td>comh-alta</td>          <td>com-halta</td></tr>
</table>

<p>
The last example is of course an abomination of the worst kind.
</p>

<hr>
<h2>Known bugs or ambiguities</h2>
<p>The word “record” is a well-known example of 
an English word where the proper hyphenation depends on 
context (verb <i>re-cord</i> vs. noun <i>rec-ord</i>).
A strict adherence to morphological hyphenation in Irish 
leads to a number of amusing (and highly-improbable) ambiguities,
many arising from the not-particularly-distinctive form
of the imperfect autonomous:
</p>
<ul>
<li><i lang="ga">bhrach-taí</i> “used to be fermented” (broad stem) vs. <i lang="ga">bhracht-aí</i> “sappiest”</li>
<li><i lang="ga">cháint-í</i> “most critical” (no hyphen) vs. <i lang="ga">cháin-tí</i> “used to be taxed”</li>
<li><i lang="ga">Cheilt-í</i> “most Celtic” (no hyphen) vs. <i lang="ga">Cheil-tí</i> “used to be concealed”</li>
<li><i lang="ga">chist-í</i> “treasures” (no hyphen) vs. <i lang="ga">chis-tí</i> “used to be handicapped”</li>
<li><i lang="ga">gcoirtí</i> “may you tan” (no hyphen) vs. <i lang="ga">gcoir-tí</i> “used to be worn out”</li>
<li><i lang="ga">chreataí</i> “used to be shaken” (no hyphen) vs. <i lang="ga">chreat-aí</i> “shakiest”</li>
<li><i lang="ga">doir-tear</i> “breeds” vs. <i lang="ga">doirt-ear</i> “spills”</li>
<li><i lang="ga">fhuad-ar</i> “bustle” vs. <i lang="ga">fhua-dar</i> “they sewed”</li>
<li><i lang="ga">fhuaf-ar</i> “odious” vs. <i lang="ga">fhua-far</i> “one will sew”</li>
<li><i lang="ga">ghais-tí</i> “used to gush” vs. <i lang="ga">ghaist-í</i> “traps” (no hyphen)</li>
<li><i lang="ga">gheal-taí</i> “used to be brightened” vs. <i lang="ga">ghealt-aí</i> “lunatics”</li>
<li><i lang="ga">ghor-taí</i> “used to be incubated” vs. <i lang="ga">ghort-aí</i> “may you injure”</li>
<li><i lang="ga">na haist-í</i> “hatches” (no hyphen) vs. <i lang="ga">na hais-tí</i> “essays”</li>
<li><i lang="ga">lé-amar</i> “we read” (past) vs. <i lang="ga">léam-ar</i> “lemur”</li>
<li><i lang="ga">luad-ar</i> “movement” vs. <i lang="ga">lua-dar</i> “they mentioned”</li>
<li><i lang="ga">meat-aí</i> “most perishable” vs. <i lang="ga">mea-taí</i> “used to waste”</li>
<li><i lang="ga">réalt-aí</i> “stars” vs. <i lang="ga">réal-taí</i> “used to be developed”</li>
<li><i lang="ga">ríf-ear</i> “reefer” (Collins Gem) vs. <i lang="ga">rí-fear</i> “will tighten”</li>
<li><i lang="ga">shá-dar</i> “they stabbed” vs. <i lang="ga">shád-ar</i> “solder”</li>
<li><i lang="ga">Shá-imis</i> “we stab” vs. <i lang="ga">Sháim-is</i> “Sami language”</li>
<li><i lang="ga">shá-faí</i> “one would stab” vs. <i lang="ga">sháf-aí</i> “ax handle” (genitive)</li>
<li><i lang="ga">speal-ta</i> “mowed” vs. <i lang="ga">spealt-a</i> “milt” (genitive, no hyphen)</li>
<li><i lang="ga">thiom-áin-tí</i> “used to be driven” vs. <i lang="ga">thiom-áint-í</i> “drives”</li>
</ul>

<p>
In the rules file, these words are placed
inside a <code>\hyphenation{}</code> statement so that TeX will not apply
the usual rule set to them.
</p>

<p>
Note there is also a potential difficulty with words like
<i lang="ga">bainte</i> which can be viewed morphologically as
<i lang="ga">bain+te</i>
(i.e. past participle) or as <i lang="ga">baint+e</i>
(genitive of a second declension noun).   The same holds
if the noun forms admits a plural, e.g.
<i lang="ga">bhaint+í</i> could also be the imperfect
autonomous <i lang="ga">bhain+tí</i>.
The current set of patterns is designed to allow
the past participle hyphenation in most cases.
Here are the other words for which this is relevant:
<i lang="ga">athoscailte, bainte, ceilte, cigilte, coigilte, cuimilte,
deighilte, déroinnte, diomailte, dúbailte, easmailte, eitilte,
fóinte, foroinnte, fuascailte, meilte, múscailte, oscailte, roinnte,
satailte, streachailte, tochailte, tomhailte, tríroinnte, tuirlingte</i>.
Other past participles have the same ambiguity “accidentally”:
<i lang="ga">ciste</i> (<i lang="ga">cist</i>=“a cyst”), <i lang="ga">coirte</i> (<i lang="ga">coirt</i>=“tree bark”),
<i lang="ga">deilte</i> (<i lang="ga">deilt</i>=“delta”), <i lang="ga">feilte</i> (<i lang="ga">feilt</i>=“felt”)
The noun <i lang="ga">cruachta</i> is a third declension example.
</p>

<p>
Finally, there are some true bugs.
They are extremely rare: as of version 1.0 
the patterns do not produce <em>any</em>
hyphen points which are not in the database
and miss just 10 out of 314,639 hyphen points. 
This is not to say that you won't discover any
bad hyphenations, but that they are the fault of
the underlying database and not of the algorithms
used to produce the patterns.
</p>